1. Field of the Invention
The present invention relates to an information processing apparatus and a locus data recognition method.
2. Description of the Related Art
A conventional device having an input unit, such as a touch panel, extracts locus data of one character input in an input frame (field) or by various character separation methods. Furthermore, such a conventional device executes character recognition in a unit of one character by using the extracted locus data as input data in the character recognition processing.
In this case, it is assumed that the locus data does not include locus data other than the one included in the recognition target character. In other words, in character recognition processing, only recognition of a character is executed. Furthermore, in gesture recognition, only recognition of a gesture is executed during gesture recognition processing.
However, a pen-input graphic editing system discussed in Japanese Patent Application Laid-Open No. 06-208654 does not consider that an editing request and a character are input at the same time. Accordingly, the conventional method separately determines whether specific input data is an editing request or a character.
In the pen-input graphic editing system discussed in Japanese Patent Application Laid-Open No. 06-208654, if the size and the position of input data is not clear, it is not certain whether the input data is determined to be an editing request or a character. In addition, in the pen-input graphic editing system discussed in the Laid-Open No. 06-208654, even if an editing request is correctly recognized, a subsequent character is likely to be misrecognized.
In addition, a data processing apparatus discussed in Japanese Patent Application Laid-Open No. 08-305821 can extract only a predetermined locus. Accordingly, the data processing apparatus discussed in the Laid-Open No. 08-305821 can correctly recognize input data only when a gesture and a character are input in a predetermined input order. But if the order of inputting a gesture or a character is not determined or used, the irregularly input gesture or character cannot be effectively recognized.
In addition, in an image recognition apparatus disclosed in U.S. Pat. No. 5,781,663, a gesture recognition unit, a character recognition unit, and a graphic recognition unit recognize input loci and merely selects a best result from a plurality of recognition results.
Therefore, the image recognition apparatus disclosed in U.S. Pat. No. 5,781,663 does not consider a method for separating a gesture and a character which are mixed in loci.
An optimal result can be output by executing recognition processing on all locus combinations by using a combination of conventional recognition units, such as a gesture recognition unit and a character recognition unit. However, in this case, it is necessary to execute recognition processing on a large number of combinations. Accordingly, the recognition cannot be completed in a practical calculation time.
In addition, in the above-described conventional method, an incorrect combination may match a dictionary pattern. In this case, misrecognition may occur. Accordingly, the above-described conventional method is not practical. Furthermore, in the above-described conventional method, a separated recognition result may be obtained by a combination in which a specific locus is included in both separated data.
If data is input as a sentence, the above-described conventional method can separate each character. However, if a gesture and a character are mixed and included in input data, a conventional method that uses a conventional separation algorithm is not appropriate and cannot separate mixed gestures and characters.
For example, if a control command parameter is written within a gesture locus having a circular shape (“∘”), the above-described conventional method cannot appropriately recognize the parameter by recognizing that the gesture locus and the parameter constitute one character.
In addition, when a control command parameter is written within a frame (field) of a circumscribed rectangle of the input gesture locus, if the input data is separated into a locus of the gesture and a part of the input character, and a locus of the other characters, the above-described conventional method cannot recognize both locus data.